Dynamic Time Warping (DTW) for Biomechanics

Functions for Dynamic Time Warping analysis including distance computation, alignment, averaging, and clustering of biomechanical waveforms. Compute DTW distance between two time series

Usage

dtwDistance(
  x,
  y,
  window_size = NULL,
  step_pattern = c("symmetric2", "symmetric1", "asymmetric"),
  normalize = TRUE
)

Arguments

x

First time series (numeric vector or matrix column).

y

Second time series (numeric vector or matrix column).

window_size

Sakoe-Chiba band width (NULL for no constraint).

step_pattern

Step pattern: "symmetric1", "symmetric2", or "asymmetric".

normalize

Logical; whether to normalize distance by path length.

Value

A list of class "dtw_result" containing:

distance

DTW distance

normalized_distance

Distance normalized by path length

path

Alignment path (matrix with columns 'index1', 'index2')

cost_matrix

Accumulated cost matrix

Details

Calculates the Dynamic Time Warping distance and alignment path between two waveforms, allowing for non-linear time alignment.

DTW finds the optimal alignment between two time series by warping the time axis. It's useful for comparing movements that may occur at different speeds or with timing differences.

References

Sakoe H, Chiba S (1978). "Dynamic programming algorithm optimization for spoken word recognition." IEEE Transactions on Acoustics, Speech, and Signal Processing, 26(1), 43-49.

See also

dtwDistanceMatrix(), dtwAverage(), dtwClustering(), dtwWarp()

Examples

# Two gait cycles with different timing
t <- seq(0, 100, length.out = 100)
x <- sin(2 * pi * t / 100) * 30
y <- sin(2 * pi * (t + 10) / 100) * 30  # Phase shifted

result <- dtwDistance(x, y)
print(result$distance)
#> [1] 48.8402